形态梯度小波降噪与S变换的齿轮故障特征抽取算法  

The morphological gradient wavelet de-noising and S transform-based gear fault feature extraction algorithm

作　　者：刘小平[1] 许桂云[1] 任世锦[2] 杨茂云[1,2] 

机构地区：[1]中国矿业大学机电工程学院,江苏徐州221116 [2]江苏师范大学计算机学院,江苏徐州221116

出　　处：《电子设计工程》2012年第22期79-82,共4页Electronic Design Engineering

基　　金：国家自然科学基金(60974056)

摘　　要：针对齿轮故障特征信号具有强噪声背景、非线性、非平稳性特点,提出采用形态梯度小波对齿轮振动信号进行降噪。首先使用形态梯度小波把齿轮振动信号分解到多个尺度上,然后对各层的细节系数进行软阈值方法降噪处理,对经过处理后的小波系数进行重构。对降噪后的齿轮振动信号采用S变换多分辨率时频分析,能够从具有良好的时频分辨率的S变换谱图提取齿轮故障特征。通过仿真试验和故障轴承的信号分析证明,该方法具有短时傅里叶变换和小波变换的优点,不存在Wigner-Ville分布的交叉干扰和负频率,能有效地提取隐含在噪声中的齿轮故障特征,适合齿轮故障的在线监测和诊断。The gear fault features are characterized with strong noise background, nonlinearity and non-stationary. The morphological gradient wavelet （MGW） is developed to eliminate noise. The gear vibration signal is decomposed by MGW, the detailed coefficients in each scale are processed using soft threshold de-nosing and the real fault signal is obtained by reconstructing the processed wavelet coefficients. Multi-resolution S transform is employed to analyze the reconstructed vibration signal, the gear fault features can be extracted from the spectral graphs generated from S transform are with good time and frequency resolution. Simulation and experiment results show that the proposed method can preserve the merits of short- time Fourier transform and wavelet transform, and does not involve the problem of negative frequency and cross frequency disturbances, which can extract the gear fault features effectively and is suitable for online gear monitoring and fault diagnosis.

关 键 词：故障诊断 齿轮 特征提取 形态梯度小波 S变换 

分 类 号：TH165[机械工程—机械制造及自动化]